A 16→64→32 MLP driving geometry in real time. No server. No API calls.
A 3D instrument you play with your voice.
Speak, hum, or move your cursor. A small neural network running on your device turns sound into geometry in under 20 milliseconds.